Quantification of swelling clays in mineral mixtures and rocks using infrared spectroscopy.

The use of a chemometric method, partial least squares (PLS) regression and the infrared technique, diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was developed for the quantification of minerals in synthetic mixtures and sandstone rocks. Particular emphasis was directed towards quantifying swelling clays since the proportion of swelling minerals present in a reservoir rock influence the composition of fluids injected into the rock to enhance oil recovery. In the first study, the capability of PLS with DRIFTS was demonstrated when low quantities of swelling mineral, smectite, were determined in synthetic mixtures with the non-swelling mineral, kaolinite, and an additional mineral, calcite. A substantial improvement in the detection limit of smectite was found using this method since lower quantities could be determined (with a maximum error of 3.8 wt%) than had been determined by previous workers using a peak fitting technique. An initial investigation of seven different types of sandstone rocks was carried out by visual inspection of XRD traces of powder and rock samples. In this work mineralogical heterogeneity was found along the length of the different sandstone cores investigated. Ideally a representative sample would be one collected from a of ball milled sample of a large portion of a core. The main programme of research was the development of a PLS model consisting of eleven mineral components, to resemble the composition of a real sandstone. This PLS model was used for the quantification of components in synthetic mineral mixtures using their DRIFTS spectra. The maximum error for predicting all the components was +/-2.8 wt%. This was in the same order as errors reported by other workers for quantification of minerals using XRD.The same PLS model was used to quantify components in seven different types of sandstone and the results have been compared with studies by independent investigators using different techniques. The PLS model predicted components in the sandstones which contained relatively high concentrations of feldspars and clays, accurately, with all components predicted within +/-2.8 wt%. PLS model was optimized in order to obtain more accurate quantification of components in the sandstones, in particular for the "cleaner" sandstones, i.e. those which contained low concentrations of feldspars and clays. Some improvement in accuracy was found. The PLS model developed in this thesis, however, was most accurate for prediction of components in sandstones which contained higher concentrations of feldspars and clays.